Beruflich Dokumente
Kultur Dokumente
EE180D Lecture 2
1.
Often, the need for control and the required use of sensors motivates the product
requirements for embedded systems
Industrial
Automotive
Infrastructure monitoring and control
Also, as the cost of sensor systems has reduced (as the result of microelectromechanical
systems (MEMS) progress), sensors now appear in consumer products
Cellular phone systems
Personal devices
Entertainment systems
Future Wireless Health devices
Sensor systems requirements may heavily constrain the design of hardware and software systems
Data acquisition requirements
Energy requirements
Signal processing requirements
2.
Physical signal provided by a source propagates through a signal channel to the sensor system
physical signal
at source
physical signal
at transducer
Signal
Channel
Transducer
Analog Signal
Processing
Analog-to-Digital
Data Conversion
Digital Signal
Processing
output signal
transducer receives the input physical signal (that may have acquired superimposed noise and
suffered distortion in the channel for a non-ideal application).
acceleration signal may result in the motion of a transducer proof mass resulting in a
displacement and a change in electronic capacitance
light intensity signal may be converted into an electronic current by a photodiode detector.
samples the analog signal and presents this to an appropriate (and optional) digital signal
processing stage
In the ideal case, sampling fidelity is achieved for signals at frequencies below the Nyquist
rate
Preferred approach
Sampling rate selection, when combined with proper filter produces varying sampling
bandwidth over which noise spectral density is integrated
Note that resolution for monitoring a signal will depend on sample rate
Many challenges exist where the combination of performance of sensor, analog and digital
systems do not permit achieving a needed resolution level.
quantization
noise
physical signal
at transducer
physical signal
at source
Signal
Channel
Distortion
Transducer
Analog Signal
Processing
noise signal(s )
Noise
Channel(s )
output signal
transducer
input -referred noise
Digital Signal
Processing
Combined
Distortion
transducer
interference signals
(cross -talk)
analog interface
input -referred noise
Sensor System
Signals
First, the input physical signal is emitted by the source and propagates through the channel.
The sensor signal channel contributes a distortion in the frequency domain and may also exhibit
nonlinear amplitude dependence in extreme examples (although most channels display linear
response to the characteristically small signals that are available to sensors).
Noise signals present at the site of the source also contribute to the input signals and these also
propagate through channels that convey noise to the transducer input. It is important to note that
noise and signal will be indistinguishable at the sensor input.
o
noise signals that arise from the physical environment are not to be considered as sensor
limitations and are to be maintained separate from evaluation of sensor systems.
Specifically, the ENS system designer must be allowed to compute a response from the sensor
system given knowledge of an input physical signal amplitude.
o
It is then the task of subsequent sensor data and information processing to separate source signals.
Now, sensor noise leads us to a separate discussion that also contributes to some aspects of
calibration, however
The most important noise source to be considered next is that of the transducer itself. Specifically,
all transducers display a noise associated with their operation.
Thus, even in the limit of a noise free environment, the transducer output is finite.
As we will see, this noise source is observed to be equivalent to some signal and this will lead
to the definition of a noise-equivalent signal level for each sensor.
Continuing to the next source of input noise the role of interference (so-called cross-talk) is
encountered.
o
Here, the transducer system is presented with signals derived from the input physical signal
of interest (for example an atmospheric pressure signal) and temperature variations that
create an interfering signal is indistinguishable from the desired signal yet arises from entirely
separate phenomena.
Finally, the analog signal processing interface also contributes an important noise source
This noise may result from the input-referred noise of the input transistor stages of the
analog signal processing system.
The design and operation of sensor platforms relies on the availability of standard definitions for
comparison of sensor systems.
sensor systems are characterized by means to avoid the impact of environmental noise, channel
distortion, or cross-talk.
The first primary sensor characteristic figure of merit is responsivity, ( a transfer function) the ratio
of output signal amplitude to input physical signal amplitude.
responsivity =
o
Sensitivity determines the frequency dependent and measurement bandwidth dependent resolution
of a complete sensing system.
o
Obtained by removing all sensor system inputs and measuring the resulting output signal
output signal results only from noise originating from the transducer itself or signal
processing section.
Since noise at the output is indistinguishable from that produced by an input signal, and
since responsivity is known or can be determined, then the value of the input-referred noise
level may be computed
This is literally the value of input signal that if applied at the input would result in the
observed noise level.
sensitivity =
frequency dependence of sensitivity must be considered since noise signal sources are
distributed in frequency.
low frequency limit of operation encounters the phenomenon of drift, a slowly changing
value of sensor output, due to low frequency noise sources.
At high frequency, a broad band, frequency-independent noise region appears due to broad
band noise at the analog signal processing interface.
4.
ENS designer must include this frequency dependence along with knowledge of desired
signal frequency spectra in computation of sensor system performance.
Next step: we will illustrate example of a typical accelerometer then consider noise sources
The combination of accelerometer and gyroscope systems offers a capability for inertial navigation.
Specifically, an inertial measurement unit (IMU) may provide rotation information with respect to an
inertial reference frame, and acceleration.
These data combined, can enable orientation of the accelerometer axes to be known. Then, double
integration of the accelerometer output can yield absolute position of the IMU in space.
Clearly, this would have immense applicability in vehicle control, wireless health, personal
localization systems, and many others.
However, fundamental drift that appears as a low frequency noise contribution, prevents accurate
inertial navigation.
o
As an example, a drift in the output of a sensor may appear as an offset signal (a near static
value of acceleration that appears in the absence of a physical acceleration input).
The figure below shows the integrated position error resulting from an acceleration offset signal
demonstrating that small offsets lead to large error in current applications
Consider typical error tolerances for an accelerometer with 1000 micro-g offset
o
Personal navigation
Vehicle navigation
The Analog Devices ADXL50 is shown below (at left) with the ADXL02 sensor element at right.
10
11
Suspension
Proof Mass
dD
Z
Inertial Reference Frame
proof mass, m,
quality factor, Q
resonant frequency, o =
k m.
D = 0 corresponding to the rest position of the proof mass at zero spring deflection
Damping will be proportional to the velocity of the mass relative to the package.
Damping force = b
dD
dt
The forces associated with the spring and damping must equal the force required to
accelerate the proof mass, so
d 2 ( Z D)
dD
d 2Z
d 2D
dD
=
kD
+
b
m
=
m
+b
+ kD
or
2
2
2
dt
dt
dt
dt
dt
12
d 2 D 0 dD
+
+ 02 D
dt 2
Q dt
a=
where, b =
m0
Q
D( ) =
This displacement serves as the signal, since it is what will be eventually converted into an electronic
signal. The responsivity (signal divided by excitement) of the accelerometer is defined by
D
=
a
a ( )
.
+ j (O ) Q
2
O
2
O
) + ( )
2
Q2
The responsivity expression simplifies for an acceleration signal frequency much less than o. In
this case,
D = a / 02
6.
and
D /a = 1/ 02
Capacitive position sensors measure variation in displacement current due to change in capacitor gap.
Suspension
D2
D1
Capacitance 1
_
_
13
This method shows high-resolution measurement capability, but, requires differential measurement
architectures to ensure linearity and low drift.
contrast to piezoresistive/piezoelectric methods one may design the system with independent control
of position sensor sensitivity and structure compliance.
Thus in both the accelerometer and gyro examples resonant frequency of the proof mass system can
be made independent of position measurement.
scaling issue
reduction in capacitance area with decreasing sensor scale results in reduction of capacitor
displacement current, thus producing reduced responsivity and reduced sensitivity.
For conventional (large scale) capacitive detectors, 10-4 / Hz sensitivity has been
demonstrated
pressure sensors
accelerometers
seismometers.
A capacitance bridge provides position measurement capability. This makes use of a switched
capacitor drive, which modulates the desired signal away from dc, reducing 1/f noise.
t
V+
V-
Switched
Capacitor
Drive
ac coupled
gain stage
output signal
Phase Sensitive
Detector
Output
Amplifier
Switched
Capacitor
Drive
Further, the feedback system results in an output always at null, eliminating gain errors. Additionally,
the force rebalance output is immune to oscillator amplitude and phase drift.
The resolution is defined by the ratio of the drive voltage to the amplifier front end noise.
Resolutions approaching 106 have been demonstrated
Stray capacitance to the substrate may limit applicability of capacitive position sensor, in general
requiring careful attention to structural design.
14
7.
Determining the length of time required for a sensor system to achieve a given level of
resolution
We will examine both transducer noise sources and electronic signal processing system noise sources
Small bias errors will thus translate into large motion/position errors.
For an ideal accelerometer, the limitation on mass reduction arises from thermally-induced mass
motion.
The thermal noise (formal analog of Johnson noise for a resistor) arises from the equipartition
theorem.
o
Specifically, absent any external influence, the total energy in each axis of oscillation
mode of a system (for example the spring mass device) is equal to kBT/2.
15
The rms value of mass position, Drms, for all frequencies, is then computed from
1
1
2
kDrms
= kB T
2
2
As an example of the magnitude of Drms, for a mass suspension of spring constant k = 1N/m, the
rms mass motion Drms 0.6 (orders larger than desired signal!).
The value of mass motion over a narrow frequency bandwidth is frequency dependent and depends
on Q. The rms value of D, induced by the thermomechanical force, FT, is determined from the
frequency dependent D value
DT =
Note that the mass motion induced by acceleration of the mass suspension, a, (the accelerometer
case or package) and the accelerometer responsivity, given above, have the same frequency
dependence. Thus, the thermal noise equivalent acceleration (TNEA), the ratio of these signals is
frequency independent
4kTO
MQ
4kTO
MQ (O2 2 ) + (O ) Q2
TNEA
M (O2 2 ) + j (O ) Q
The integral of (1/2)|DT|2 over all frequencies determines the value of FT. The frequency
dependent mass motion in a 1Hz bandwidth is,
DT =
FT
We will analyze an automotive collision impact microsensor example and observe that this
noise represents an important sensitivity limitation for this application.
16
Gyroscope Applications
o
Braking control
Sensitivity
o
Advantages
o
Compact, low-power
Disadvantages
o
Reduced sensitivity
Increased drift
Rate Gyroscope converts rotation rate into a displacement. that may be measured directly, using
the Coriolis Force.
The Coriolis Force is an apparent force observed in a rotating reference frame that appears to act
on moving objects.
17
An Ideal Rotation Rate Transducer may be implemented using the Coriolis Force principle.
Proof Mass
Suspension
Harmonic
Coriolis
Deflection
Harmonic
Horizontal Mass
Motion
The Coriolis force operates on moving masses and is harmonic with drive frequency
R
rA
rA
The rate gyroscope system will carry two masses (m) supported on structure having vertical
restoring forces of k, quality factor Q, resonance frequency, d. Masses supported at radius R
with motion along the x-axis of
r1 = R + rA sin d t
r2 = R rA sin d t
Masses are, therefore, being driven at the resonance frequency of the vertical suspension. This
provides the greatest natural amplification of out-of-plane Coriolis force-induced motion.
18
FC = 2mV
Consider Z-axis out of page, X-axis along line of motion, Y-axis vertical direction on drawing.
For the case of rotation about the Z axis, the deflection of the mass by the Coriolis force is
FC = 2mrA d cos d t
y =
FC Q
FC Q
=
= 2rAQ
2
k
(m d )
d
cos d t
Deflection is enhanced by driving the oscillator system at its resonance with a gain of Q.
For Earth rate = 4.17 x 10-3 degrees/sec = 7.3 x 10-5 rad/sec, rm = 10, Q = 103,
d = 104 rad/sec,
yC = 1.46 x 10-10 m
Small deflection response introduces error for devices operating with high 1/f noise. The
application of feedback bridge methods will be critical.
Conclusion : ideal rate gyro concept should meet inertial navigation requirements with
substantial margin.
Bias errors and output drift, limiting sensitivity, must arise from non-ideal operation - focus of
current development.
The use of a rate gyroscope for measuring the orientation of an object encounters error.
Since the rate gyroscope measures rotation rate (rather than orientation) then orientation is
obtained by sampling the output of the rate gyroscope and integrating the rate gyroscope output
with respect to time, once.
19
Also, small offset errors in rate measurement are integrated and lead to an accumulation of error.
The figure below shows the integrated position error resulting from a rotation rate offset signal
demonstrating that small offsets lead to large error in current applications
Consider typical error tolerances for a rate gyroscope with a 1 milli-degree/second offset
o
Personal navigation
20
Design Goals
o
High responsivity
High Q
Mode matching
Design Challenges
o
Material properties
Mode matching
Drift
Solutions:
o
Feedback control
Current Implementation:
o
21
From A Symmetrical and Decoupled Microgyroscope with electroforming process on insulating substrate
Alper, S.E., Akin, T., Proceedings of The 16th European Conference on Solid-State Transducers, September
15-18, 2002
22
Triaxial Rate Gyroscope Proof Mass, Actuator, and Sense Structures for the STMicroelectronics
L3G4200D Top View
23
Triaxial Rate Gyroscope Proof Mass, Actuator, and Sense Structures for the STMicroelectronics
L3G4200D Perspective View
24
Triaxial Rate Gyroscope Sense Electrode Structure underlying proof mass system
25
Measurement axes
Consider X axis rotation (vertical and in plane of page for figures above).
Now, masses at left and right in the figures above move to left and right in
differential motion
Consider Y axis rotation (horizontal and in plane of page for figures above).
Now, upper and lower masses in the figures above move to upwards and
downwards in differential motion
Now, consider Z axis rotation (vertical and perpendicular to the plane of the page for
figures above).
Now, masses at left and right in the figures above move to left and right in
differential motion
Note the very much larger area associated with these electrodes as required to
establish sufficient capacitance to ensure sensitivity values.
26
Responsivity does not depend on sensor scale. Mass and R do not appear. Contrast with system
consisting of accelerometers where signal vanishes as moment arm and mass go to zero.
Sensitivity does depend on sensor scale. Thermomechanical noise, below, represents the ultimate
limit on mass reduction.
Several error sources - example for non-planar/parallel motion, equal masses, torsional
unbalance force. Consider case of simple torsion mounted tuning fork gyro.
rA
rA
Torque
Measurement
Fa = md2sindt
But, a component of this force, R/R, acts in the y-direction to creat a torque about Z-axis.
Now, this drive force is 900 out of phase with Coriolis force. Thus, the y-axis deflection signal
created by this error force may be partially rejected by phase sensitive detection. The ratio of the
magnitudes of these two forces is
FC
FA
R d
R
27
For a system with the parameters above, and earth rotation rate, the signal/error force ratio is
one for R/R tolerance of less than 2 x 10-8. The introduction of an accurate
phase
sensitive detector may relax this requirement to 10-7.
Clearly, this error source accounts for a substantial bias error - and, therefore, drift.
Torsional unbalance misalignment may arise from tolerance of fabrication process and
from out-of-plane forces of input axis drive mechanism.
Phase/amplitude imbalance
rA
rA
Torque
Measurement
rA
rA
Torque
Measurement
o
Lateral acceleration error - unequal mass motion excited by lateral acceleration - appears
as signal
Elastic modulus drift error - torsion suspension properties scale output value and
determine bias value. Temperature coefficient of suspension determines system
properties.
28
Data sheet for device is below. Note that responsivity is termed sensitivity and sensitivity is termed rate
noise density.
Responsivity ranges
o
o
Sensitivity
o Rate Noise Density
o At a measurement bandwidth of 50 Hz: 0.03 degrees/sec-(Hz)1/2
o Typical uncertainty in 1 Hz bandwidth is 0.03 degrees per second rms
o Typical uncertainty in 50 Hz bandwidth is 0.2 degrees per second rms
29
10.
Engineering design goals will seek to reduce system cost (and reduce performance) until
sensitivity requirements are just satisfied with no excess sensitivity
Thus, sensor systems will generally be found operating at or near their theoretical limits
Sample rate
Start with Parsevals relation the integrated power (squared voltage) in a signal over all time is
equivalent to an integral over all frequency of the squared modulus of the Fourier transformed
voltage.
v 2 (t) dt = V ( f )V * ( f )df
V( f ) =
v(t)e
j 2 ft
dt
Now,
o
PSD( f ) = V ( f )
Consider a band limited signal with signal being zero outside the band of f1 < f < f2
Can define a root mean square (rms) signal that is conveniently observed by an instrument
30
Assuming a signal that is continuous in time produces an rms signal that is stationary
vrms =
f2
PSD( f )df
f1
Without proceeding further we can draw an important conclusion about system design
o
Consider further in the usual case that the design specification requires establishing a
maximum threshold value for rms amplitude
Then note that this relationship implies that for a reduction of noise by a factor of requires
a decreased in bandwidth of 2
PSD may be measured at a sample rate other than 1 Hz and referred to a 1 Hz bandwidth
(see Chapter 4 of reference text)
Now, for example accelerometer, compute rms acceleration noise based on fundamental limit
For this,
PSDTNEA
And,
a RMS =
f2
PSD( f )df
f1
4kT O
MQ
f2
4kT O
df =
MQ
f1
f 2 f1
4kT O
MQ
For the automotive example, the accelerometer that provides a solution is a silicon micromachined
devices, for example the Analog Devices ADXL series
31
This is significant since even collision detection must occur early in collision cycle when acceleration
values are small.
This noise signal scales with square root of bandwidth for frequency independent noise
32
NOISE SOURCES
One of the most important sources of noise (and reductions in sensitivity) result from electronic
noise sources
o
It is very important to note that a fundamental relationship exists between sensor energy
requirements and sensitivity
This must be understood in order to establish a properly optimized (or even feasible) system
design
Electronic noise appears from a wide range of sources. Some are fundamental to measurement
processes. These include
o
Flicker or 1/f noise: A fundamental noise source arising from phenomena including the
population and depopulation of individual charge carrier traps in semiconductor devices.
1/f noise appears in a broad range of disciplines and appears to be a consequence of
ubiquitous switching phenomena.
Noise impacts system design through both sensitivity and the power required to achieve
sensitivity.
VJN = 4k B TRB
o
To provide an example of its magnitude, for R = 1M, VJN = 128 nanovolt/(Hz)1/2. Johnson
noise is important to include in analysis of high-sensitivity piezoresistive transducers, bolometers,
and a wide variety of other devices
This fundamental noise source represents a limit for devices ranging from disk drive read heads
to wireless receiver systems.
33
Electron shot noise is observed for current flow I over (or through) an energy barrier in a pn
junction, Schottky barrier, or tunneling device.
o
ISN = 2eIB
where e is 1.6 x 10-19 C. Again, bandwidth is often neglected so that shot noise is stated in
units of A/(Hz)1/2.
Photon shot noise may limit optical sensors, in particular CCD and CMOS active pixel
sensors.
The largest source of noise at the transducer/amplifier interface is typically transistor amplifier front
end noise, since signals at this stage are of lowest amplitude and the effects of noise are greatest.
o
Bipolar transistor devices show the lowest input noise at low and high frequency for any typical
semiconductor device.
o
Effects of transistor noise in stages following the first amplifier are reduced due to the
amplification gain. Thus it is the design of the first amplifier that is critical.
Often, however, the properties of these devices may not be exploited since bipolar devices
also present a low input impedance device and so are not compatible with a number of
important transducer systems.
Low frequency performance is limited by 1/f noise. In the bipolar transistor, 1/f noise appears as
both a voltage and current signal. The voltage and current noises are
Vn = 2efC IB rb2
In = 2efC IB
1
f
1
f
where fC is the corner frequency, is in the range (1, 2) depending on the device, 1, rb
is the base resistance and IB is the base current.
Noise increases below the corner frequency, impacting low frequency sensor stability.
34
Since both voltage and current noise are present, then an optimization must be performed to
select the appropriate value of system input resistance to ensure that these separate
contributions result in a minimum noise.
The input noise increases in limits of both large and small input resistance.
Assuming a corner frequency of 50 Hz, =1, Ib=6.25 x 10-5 A, rB=1 k and =1,
The MOS field effect transistor (FET) is a high input impedance device, and consequently is
compatible with a wide variety of high output resistance transducer systems.
o
Characteristics are low input noise at high frequency and 1/f noise.
Origins of 1/f noise are attributed to charge carrier trapping at interfaces and in the gate
insulator. The MOS field effect transistor 1/f noise is
Vn =
K
(WL)C ox
1
f
for an MOS device of width W, length L, gate capacitance per unit area, Cox , and flicker
noise coefficient, K.
Input-referred, frequency-independent Johnson noise appears also in transistor devices and scales
with transistor transconductance and base resistance. Its value is:
Vn = 4k BT rb +
2 g m
35
Vn = 4k BT
3g m
In FETs, noise decreases with increasing gm. gm increases with drain current as (ID)1/2. Thus the
input stage noise decreases with increasing power dissipation. The noise also decreases with
increasing device area.
Critical result:
Large increase in drain current and power required to meet sensitivity demand
Where possible, low frequency noise may be removed through a heterodyning approach
Prior to being supplied to an input stage, an input signal may be modulated (or chopped)
with a switching mechanism that preserves signal amplitude properly.
Then, this high frequency signal may be supplied to a preamp such that its new frequency
spectrum lies above the 1/f corner frequency
The ubiquitous problem of 1/f noise may be eliminated under restricted conditions.
modulates an input low frequency signal and effectively translates this signal to high
frequency where amplification may occur in the absence of 1/f noise.
Example: non-overlapping clock phases switch the input, ac signal at the clock
frequency producing a carrier with amplitude Vi at the output.
36
output signal
Phase Sensitive
Detector
Output
Amplifier
1
2
This technique allows the subsequent use of ac amplifier stages that now operate on this high
frequency signal.
_
A phase sensitive detector then demodulates the ac signal at the frequency and phase of the
modulating clock, rejecting noise at all other frequencies.
Signal chopping is important in a variety of sensing systems, including optical detectors (which
include a light chopper/modulator) and the rate gyroscope (rotation rate provides high-frequency
signal) In a number of other sensors the input parameters may also be modulated.
Direct benefits of signal chopping result from replacing the sensitivity that may have been limited by
low frequency 1/f contributions with much lower amplitude broadband noise. This may result in
several orders of magnitude improvement in sensitivity.
37
12.
Responsivity (V/m/s2 ) =
2
O
) + ( )
2 2
Q2
f
vn ( f ) = Vn 1 + c
1/2
(V /
Hz
where vn is a noise density and Vn = 100 nanovolt/ Hz and fc = 100 Hz and with o = 2 x 2000 rad/sec and
Q = 1. The output signal from this accelerometer will include the product of input signal acceleration and
responsivity summed with this voltage noise. These values are typical of the first automotive collision
detection sensors and are typical of a device intended for 1 kHz bandwidth operation.
38
39
Now, let us consider a new system where sensitivity improvement is sought. Resonance frequency may be
reduced from 1000 to 100 Hz.
40
Finally, sensor system sensitivity has this form for 100 Hz resonance frequency
41
Sensor device energy dissipation is divided into transducer demands and the demands associated with
analog front end, digital data conversion, digital signal processing (as required) and feedback control
systems (as required).
o Note that many sensors incorporate feedback control and include actuation. Virtually all
inertial navigation sensors, for example, include this.
Noise present in one spectral region is of much larger amplitude than signal present
in other spectral regions
Sample rates may only exceed that of the Nyquist rate for atmospheric variables that
may be less than 1 Hz
F=
Example:
F 6 x 109 (Hz/W)
42
But, noise scales approximately as the fourth root of drain current (in strong inversion
mode) operation for a fixed design.
Reducing this input referred noise to 10nV/(Hz)1/2 may require power supply current
greater than 1mA.
43
14.
It is useful to consider sensor types based on transducers available that require low or even
no energy for their operation and may be coupled to low power data conversion systems.
Estimated energy demand for data conversion based on current state-of-the-art systems are
included.
Typical
Transducer
Sampling
Bias Power
Resolution
Analog Front
End and Data
Converter
Power
Sensor
Transducer
Mechanism
Typical Sampling
Rate
Temperature
Thermistor
0.01 0.1 Hz
10b
100W
10 - 100W
Temperature
Thermocouple
0.01 0.1 Hz
10b
100 1000W
Humidity
Capacitive
(Vaisalla)
0.001 0.01 Hz
10b
1mW
10 100W
Photodiode
0.01 100Hz
10b
10 100W
Infrared Single
Pixel Motion
Pyroelectric
0.01 100Hz
10b
10 100W
Pressure
Capacitive
0.001 0.01 Hz
10b
1mW
10 100W
Strain
Piezoresistive
0.001 1 Hz
12 24b
100W
1 300mW
Vibration
Piezoelectric
0.001 1 MHz Hz
10b
10W 100mW
Vibration
Geophone
0.01 1000Hz
16 24b
10W 300mW
Microphone
Electret
10Hz 20 kHz
12 18b
10W 300mW
Acceleration
Accelerometer
DC 10kHz
10 18b
0.1 100mW
0.1 10mW
Visible Image
CMOS APS
1 30 fps
10b
10mW
10 100mW
Infrared Thermal
Imager
Bolometer Array
1 10 fps
10b
1W
500mW
Visible Imager
High resolution
CCD
1 30 fps
24b
1W
1W
44
Structure
1. suspended proof mass
2. proof mass and magnet combined
3. coil surrounds magnet/mass
Operation
proof
mass
magnet
flexure
coil
flexure
coil
proof
mass
magnet
45
Preamplifer input referred noise must not exceed this sensor noise (to achieve optimal
sensitivity)
Time synchronization
Processor
ARM 9
ADC
Architecture
46
16.
Conducting polymer
o
Composite material
Insulating matrix
1. Low elastic modulus
17.
Nutrition biomarkers
Blood oxygen saturation sensors provide a non-invasive, low cost, accurate measurement system.
Principles:
o
47
1000000
100000
HbO2
Hb
10000
1000
100
250
450
650
850
Wavelength (nm)
Optical absorption spectroscopy shows that absorption of oxy-hemoglobin and deoxyhemoglobin is equal near 800 nm and at longer wavelength
Optical transmittance, T, (for large transmittance and small absorbance) follows Beers Law:
T = 10 cd
where is the absorption coefficient (per mole per cm3), c is the concentration in moles per cm3
and d is the optical path length in the medium
First, this can be designed for measurement of transmission of light through tissue (for
example a finger tip).
Now light propagates through tissue and encounters tissue containing no blood, or
venous blood (deoxygenated hemoglobin blood returning in veins to the heart), and
arterial blood (oxygenated hemoglobin blood arriving from the heart).
Arterial blood arrives with a time varying pressure that follows cardiac pump activity.
Arterial blood volume fluctuates in synchronization with pulse activity.
48
(near infrared devices) at 840 nm for detection of both oxy- and deoxyhemoglobin
Two LED sources used (operating modulated in time) and a single detector.
49